Skip to content

[WIP] Introduce additional_labels #306

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Closed
wants to merge 83 commits into from
Closed

[WIP] Introduce additional_labels #306

wants to merge 83 commits into from

Conversation

AntonioMirarchi
Copy link
Contributor

@AntonioMirarchi AntonioMirarchi commented Mar 15, 2024 •
edited
Loading

This PR proposes several enhancements to enable the usage of extra arguments in both training and inference. The key modifications include

  • standardizing the naming conventions for all extra fields across different data loaders;
  • removing q and s as explicit args in the forward;
  • introducing additional labels through a new dictionary for model initialization, example:
additional_labels:
     tensornet_q:
        label: total_charge
        learnable: true
        init_value: 0.1
  • add TensornetQ nn.Module
  • extra args expansion from molecule-wise to atom-wise inside the model
  • tensornet_q tested with total_charge and partial_charges, reproducibility achieved
  • rewriting some tests according to the new methods
  • allowing multiple additional_methods implementation
    At the moment only 'tensornet_q' supported as additional_methods but more can be added

guillemsimeon and others added 30 commits March 5, 2024 11:42
@guillemsimeon
Update tensornet.py
43d8f9f
@guillemsimeon
Update model.py
9ccc2c0
@guillemsimeon
Update model.py
25854fb
@AntonioMirarchi
move to extra_fields implementation
cc6de7a
@AntonioMirarchi
remove charge and spin, this will go to extra_fields
9478a92
@AntonioMirarchi
force hdf5 dataset to common-unique data structure
c7014e7
@AntonioMirarchi
force ace dataset to common-unique data structure
1f4928c
@AntonioMirarchi
remove charge and spin flag from train.py
7f238dc
@AntonioMirarchi
add extra_fields to argparse
3250fc4
@AntonioMirarchi
memdataset to common-unique data-structure
736af54
@AntonioMirarchi
qm9q to common-unique data structure
d495044
@AntonioMirarchi
ET to extra_fields
014a06c
@AntonioMirarchi
add extra_fields documentation to the ET
b081a09
@AntonioMirarchi
transformer to extra_fields
2910694
@AntonioMirarchi
small fix in ET documentation
c0489af
@AntonioMirarchi
graph-network to extra_fields
cb2229a
@AntonioMirarchi
remove optional tensor for extra args, it's needed by default
2fe05bc
@AntonioMirarchi
remove extra_args from model forward, extra_fields_args it's only needed
3150672
@AntonioMirarchi
add self.extra_fields to architectures
6055357
@AntonioMirarchi
remove all 'q' specific function, to move to more general extra_fields
e3ebbbe
@AntonioMirarchi
change variable name t additional_labels and allow to be only a dict
ea3acd6
@AntonioMirarchi
remove architectural redundancy
c34fc52
@AntonioMirarchi
move to additional_labels verion
32f2456
@AntonioMirarchi
use extra_args
ef4cfbe
@AntonioMirarchi
tnsnet v2 with tensornetQ class as additional method
ebb08c2
@AntonioMirarchi
update warning message
91c7cde
@AntonioMirarchi
force labels to be atom_wise
fbfafaf
@AntonioMirarchi
remove unused arg
088dd5f
@AntonioMirarchi
fix extra_args_nnp generation
9479ae7
@AntonioMirarchi
remove old code residue
cd8ab2d
@AntonioMirarchi
remove old comment
052261e
@AntonioMirarchi
update to get the additional labels from argparse as discussed in the PR
c0902c5
@AntonioMirarchi
fix documentation, include extra_args in the model's input
917cb7a
@AntonioMirarchi
to black
452362b
@AntonioMirarchi
remove Any from typyng import
cfd0f53
@AntonioMirarchi
initialize_additional_method as free standing function
5e02fc7
@AntonioMirarchi
remove Any from typing import because not used
3470f7e
@AntonioMirarchi
remove Any from typing import
2ec850a
@AntonioMirarchi
update for loop in the forward step and extra_args/static_shapes asse…
781a3a7 
…rtion
@AntonioMirarchi
send always extra_args
1a30039
@AntonioMirarchi
add reset_parameters for tensornetQ
e811a5e
@AntonioMirarchi
add an assertion to verify that additional_labels is not specified fo…
27f13db 
…r models that do not implement it
@AntonioMirarchi
reintroduce charge and spin for backward compatibility
3ee76de
@AntonioMirarchi
to black
d5a9e5f
@AntonioMirarchi
remove comment
75c6726
@AntonioMirarchi
let tensornetq's operations more efficient
591253c
@AntonioMirarchi
move to label_callbacks instead of additional_methods
4cd6313
@AntonioMirarchi
rename to additional_labels_handler
6b72a00
@AntonioMirarchi
get torch.jit compatibility, comprehension ifs are not supported yet
fc58fb3
@AntonioMirarchi
remove unused import
0ac63d9
@AntonioMirarchi
use warning instead of assertion
8a4a181
@AntonioMirarchi
more readable format
fdf12aa
@AntonioMirarchi
update test_model considering extra_args will be always passed to the…
580cd34 
… forward
@AntonioMirarchi
move prefactor and tensornet_q operation from the interacton layer to…
d483230 
… tensornet nn.module
@AntonioMirarchi
fix typo
9e546dd
@RaulPPelaez RaulPPelaez mentioned this pull request Mar 19, 2024
Merged
@AntonioMirarchi AntonioMirarchi mentioned this pull request Jan 27, 2025
Open
@AntonioMirarchi
Copy link
Contributor Author

Closing this PR for now as an alternative, more streamlined solution has been proposed in #353.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@RaulPPelaez RaulPPelaez RaulPPelaez left review comments

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@AntonioMirarchi @RaulPPelaez @guillemsimeon